Probably in "visuospatial thinking".
I assume here that you are asking about the modality of thinking. This is not a well studied area in cognitive science. I believe the question stems from an underlying assumption that most people think in their primary language (or possibly switch if they are fully bilingual). This was certainly the view early on in the history of psychology.
In fact, the earliest theories about the development of "inner speech" suggested that it develops around age 8 when self-talk or "private speech" - basically young children talk to themselves a lot - is gradually replaced by inner speech (children are often encouraged to use their "inside voice"). This seems to imply that prior to this age, children don't have the "internal monologue" that adults are familiar with.
Later on, "visual thinking" was examined, primarily in the context of determining and catering to "learning styles". In this context, several different kinds of visuospatial thinking styles have been considered, including picture thinking, shapes and transformation, and spatio-temporal thinking. The possibility of kinesthetic or tactile learning was also identified. The concept of learning styles is largely discredited now, though it did call attention to the special needs of children with learning disabilities such as dyslexia and autism, where visual thinking may be more common.
The more modern view is that while people may have a dominant mode of thinking, they switch between modalities depending on context - such as the nature of a problem that they are trying to solve. In general, we can only really learn about thinking modalities from self-reports of introspection, which is not considered to be very reliable. To improve reliability and validity, a formalized scale should be developed - this has been attempted in the past, but is not in common use.
To learn more about what pre-verbal thinking might be like, we could try asking children how they thought prior to developing inner speech, but this requires a skill called "metacognition" - the ability to think about thinking - that is not well developed in children. It also seems that adults no longer remember what their pre-verbal thinking was like.
Feral children do seem like a possible opportunity to get around this. However, in practice, feral children are (thankfully) extremely uncommon, and are either discovered and taught language early enough, or in the very rare cases (case?) where it may be too late, we don't have a way to communicate to even ask the question.
A possibly better opportunity is deaf children. Most deaf children are not born to deaf parents, and hence are not exposed to language (sign language) as early as hearing children. In a few rare cases, deaf children learn sign language as their first language as late as adolescence. However, even in these cases, it is unlikely that these children did not have prior exposure to at least some language concepts, as they communicate with their family and friends, and may learn to read, lip-read, and gesture.
Deaf people often report thinking visually - in sign, print, or lip-reading for example. Blind people to some degree, and deafblind people to a greater degree report some tactile thinking - presumably if they use braille or a tactile sign language. Thinking modality appears to match sensory modality and perhaps have a preference for language modality, but may still switch modalities depending on context. Apart from these indirect hints however, we don't have much insight into the thinking of adults without language.
Language and thought:
Confounding matters further, there are also a variety of different theories about the relationship between language and thought that may have some relevance to the question. For example, one influential theory suggests that language is essential for thought, that language shapes thought, and that thought without language is limited in that certain cognitive faculties cannot exist. Empirical evidence supports some interpretations of this theory and not others.
Another approach suggests that thought has its own "language" - ie, that people think in a private language that is independent of their language acquisition - variously called "mentalese", "i-language", or the "universal grammar". Although this view has much support in linguistics, it has also met some criticism.
Yet another view (tied to the private speech theory mentioned above) suggests that thought is language - a form of intrapersonal communication that is used for deliberation, speech rehearsal, and possibly stress relief for example. Other scholars emphasize the relationship between thought and perception, suggesting that thought modality should reflect sensory modality.
The bottom line is that even the definitions of language and thought are under question, so it's difficult to say whether people without language think in any way comparable to verbal thought, or whether they think the same way that we do without the accompanying narrative, or whether they think in a way that matches their (visuospatial) perceptual experience.